Soft Sensor
A soft sensor, or virtual sensor, is a software model that estimates a process variable in real time from other measurements instead of measuring it directly. Using mathematical or machine learning models, soft sensors replace or complement physical sensors in industrial and manufacturing control and monitoring, offering cost-effective, adaptable virtual measurements.
A soft sensor is built from historical data that pairs easy-to-measure signals, such as temperatures, pressures, and flow rates, with the hard-to-measure value of interest, often taken from lab samples. The model learns the relationship and then outputs a continuous estimate from live readings.
For example, a food or chemical plant might estimate product moisture or viscosity every few seconds from oven and mixer readings, rather than waiting hours for a lab test. Operators can adjust the process immediately instead of after a batch is finished.
A soft sensor differs from a physical sensor because it calculates a value instead of measuring it. Its estimates are only as good as its model, so it is periodically checked against lab or physical measurements and retrained when the process changes.